Beginning with the successful exfoliation of monolayer graphene, two-dimensional (2D) materials, including h-BN, transition metal dichalcogenides (TMDs), Xenes and MXenes, have attracted significant interest in the past dozen years due to their unique physical and chemical properties. As a specific Xene branch, 2D atomic sheets of group VA elements possess tunable electrical, thermal, optical properties compared to their bulk state, thus having a broad prospect in the field of CMOS applications, such as field-effect transistors (FET), sensors, optoelectronics, thermoelectrics, topological insulators and so on. Group VA consists of nitrogen, phosphorus, arsenic, antimony and bismuth, all of which have a puckered or buckled layered structure except for nitrogen. Black phosphorus (BP) is the most thermodynamically stable allotrope of phosphorus under standard conditions^[1], with puckered orthorhombic structure formed by a hexagonal ring, namely α phase (space group Cmca), while blue phosphorus possesses a buckled rhombohedral structure^{[2, 3]}, known as β phase (space group R m) as Fig. 1(a). For As, Sb and Bi, the most stable structure is β phase^[4], which are also called gray arsenic, gray antimony and metallic bismuth.

As the most intensively studied 2D Xene material, BP is known for its tunable direct band gap (0.3 eV for bulk and ~2 eV for monolayer), anisotropic transport properties, thickness-dependent anisotropic optical response^[5], and high carrier mobility (of order 10⁴ cm²V^–1s^–1 along the zigzag direction)^[6], promising for nano-electronic and optoelectronic devices. At room temperature, BP FET shows a thickness-dependent mobility up to 10³ cm²V^–1s^–1, and the drain current modulation is on the order of 10⁵, which demonstrates the potential for CMOS device applications^[7]. Nevertheless, poor air stability^[8] is an obstacle for practical applications of BP, although air-stable BP devices with encapsulation have been demonstrated^[9].

2D monolayer arsenene possesses carrier mobility as high as 635 cm²V^–1s^–1 for electrons and 1700 cm²V^–1 s^–1for holes as calculated^[10]. A large indirect bandgap of ~2.49 eV for β-arsenene is also predicted^[1]. Nevertheless, Pumera et al.^[11] reported that arsenic nanosheets were considered to be nonnegligible highly toxic, which may restrict their biomedical and related applications. Monolayer β-antimonene is calculated to have an intrinsic indirect bandgap of ~2.28 eV^{[1, 12]}, and high mobility similar to arsenene^[10]. Superadd good air stability, high surface activity, moderate toxicity and topological edge states, 2D antimony shows great potential in battery, topological insulators, field-effect transistors, electrocatalysis and so on^{[8, 11, 13-15]}.

As the last element of the VA group, β-bismuth possesses the same rhombohedral structure as arsenic and antimony^{[1, 4]}. Due to the heavy atom, bulk bismuth has a strong spin-orbit coupling effect^{[16, 17]}. And p bands cross the Fermi level at T and L points as Fig. 1(c) presented, leading to a small overlap of about 30–40 meV in the valence band and conduction band^{[18, 19]} as shown in Fig. 1(d). Hence, bulk bismuth displays metallic behavior, known as metallic bismuth. 2D bismuth exhibits a lower carrier concentration than that of normal metals^[20], long Fermi wavelength results in a semimetal to semiconductor transition when the thickness is down to ~30 nm on account of the quantum size effect^[21-23]. Unlike BP, the scanning electron microscope (SEM), X-ray diffraction (XRD) and Raman characterizations of vapor-deposited bismuth films did not show significant degradation after 30 days^{[24, 25]}, demonstrating their good air stability. At room temperature, 0.1–2 μm bismuth thin films show high carrier mobility up to ~2 × 10⁴ cm²V^–1s^–1[26]. The small carrier effective mass (~0.001m_o) and the long mean free path can produce giant magnetoresistance effect in 2D bismuth^{[20, 23, 27]}. Beyond that, low thermal conductivity, optical anisotropy, high catalytic activity^[28], high specific capacity^[29] and mild cytotoxicity^[11] make it a promising candidate for future nanoelectronics.

With the rapid increase in 2D materials research, many emerging 2D materials such as graphene, BP and TMDs, have been demonstrated to be applicable to CMOS devices, including transistors, memories and inverters, benefiting from their high carrier mobility or tunable bandgap^[30-33]. In general, 2D bismuth’s outstanding properties in high mobility and a suitable bandgap will also qualify it for CMOS applications.

Starting from the structure and properties of bismuth, this review summarizes various physical vapor deposition of 2D bismuth, followed by its existing or potential CMOS applications. Besides, bismuth derivatives are mentioned as a strategy to enhance the performance of 2D bismuth. Finally, a brief summary on the current challenges and future prospects are discussed. Besides, it should be noted that in this review, “bismuthene” refers to bismuth from monolayer to 4 nm, “2D bismuth” indicates 4–30 nm (critical thickness for observing quantum confinement effect^[20]), and “thin film” is for 30 nm or thicker bismuth.

As a cousin of phosphorene, the unique structure of bismuthene gives rise to extraordinary electronic properties. To utilize the properties in potential applications, it is necessary to understand the relationship between the atomic, band structure and electronic properties of 2D bismuth.

Bismuth possesses a rhombohedral A7-type structure, namely β phase as shown in Fig. 1(a), space group R m that each Bi atom is covalently bonded to three nearest-neighbor atoms and has and three next-nearest neighbors slightly further away^[34]. For 2D bismuth, β-phase Bi (111) and α-phase Bi (110) are two distinctive allotropes as shown in Fig. 1(b) with similar average binding energy according to the same simulation method^[4], both of which can be realized experimentally as seen in Section 3.1. Note that Bi (111) and (110) planes in the rhombohedral coordinates are equal to (003) and (012) in the hexagonal coordinates^[20]. Bi (111) and (110) are reported to be topologically non-trivial and possess interesting electronic properties^[35-37]. The interlayer distance is ~0.328 nm for Bi (110) planes and ~0.395 nm for Bi (111) planes according to PDF#44-1246. The lattice parameters of Bi crystal are a= b= 4.547 Å, c= 11.8616 Å, α= β= 90°, γ= 120°, bond length d= 3.07 Å, buckling parameter ΔZ= 1.60 Å. For monolayer β-bismuthene, the optimized lattice constant will slightly decrease to ~4.36 Å, and the buckling parameter will increase to ~1.73 Å^{[38, 39]}, indicating that low dimensionality has an effect on atomic structure parameters^[40]. Interestingly, there is a transition from puckered (similar to BP) to buckled (similar to silicon) atomic structure once bismuthene grows thicker than 6 monolayers during the epitaxy process^[41], which will be introduced later on.

Bismuth has an electron configuration of 6s² 6p^3[20]. Owing to the shallow electron and hole pockets at the L and T points of the 1st Brillouin zone in bismuth, charge carriers exhibit small effective masses, a long Fermi wavelength (~30 nm), a low carrier density (3 × 10¹⁷ cm^–3) as well as metallic behavior. It is predictable that when the thickness of 2D bismuth decreases to tens of nanometers, quantum size effect will become significant and result in fascinating properties. To be exact, when the confinement produces enough energy to lift the overlap between valence band and conduction band, a semimetal to semiconductor transition will occur at thickness ~30 nm as experimental measurement^{[20, 22, 23]}.

In general, monolayer bismuthene possesses a moderate direct bandgap ranging from 0.43 to 0.99 eV depending on different computation methods^{[1, 39, 43-46]}, whereas bulk bismuth shows a small band overlap of 30–40 meV. Fig. 2(a) illustrates the band structure of bismuthene with or without spin-orbit coupling (SOC), and shows that strong SOC will narrow the bandgap^[39]. Actually, a large topological bandgap of ~0.8 eV for monolayer bismuthene on the SiC substrate has been reported^[17]. Narrow gaps of 0.4–0.5 eV are demonstrated in bismuth nanostructures including nanoflakes, nanoribbons and nanorods^{[47, 48]}.

Strong SOC effect leads to a splitting of the states in heavy Bi atoms, and the large energy difference between the 6p_1/2 and 6p_3/2 is 1.5 eV after splitting^[20]. However, the spin degeneracy cannot be lifted in the band structure because of the inversion symmetry. As for 2D bismuth, the spin-orbit splitting will strongly affect the band structure and Fermi surface due to the broken inversion symmetry, which is different from bulk bismuth with inversion symmetry. Koroteev et al. showed that the spin-orbit interaction results in a strong splitting of the surface state bands and changes the surface-state dispersion and the corresponding Fermi surfaces on low-index surfaces of Bi^[16]. Fig. 2(b) shows the electronic structure of Bi (111) together with the bulk band structure for the (111) surface calculated with or without SOC. Angle-resolved photoemission spectroscopy (ARPES) also agrees with the calculated strong splitting with SOC as shown in Fig. 2(c). These intriguing characteristics make 2D bismuth a promising spintronic material.

Material synthesis or growth is the first step to perform experimental study on the aforementioned exotic properties of bismuthene. To date, a number of methods have been developed to fabricate 2D bismuth, including physical vapor deposition, liquid exfoliation^{[11, 14, 29, 49]}, wet chemical synthesis^[50-52] and so on. Different preparation methods have diverse impact on thickness, size, morphology, structure and properties of 2D bismuth. Recently, Liu et al. reviewed 2D bismuth preparation methods including liquid exfoliation and wet chemical methods^[42]. Liquid exfoliation is a mature fabrication technique to obtain group VA nanosheets^[4] from bulk with the assistance of sonication or aqueous shear. Wet chemical methods use a Bi-based compound as the raw material to obtain 2D bismuth through chemical reactions. The products of the two are usually small pieces and are suitable for battery materials, catalysis, biosensing and so on^{[14, 29, 50, 52]}. Thin films applied to CMOS need to be large scale, continuous and uniform. Generally speaking, physical vapor deposition methods could yield large area pristine films with controllable scale on desired substrates, which is compatible with CMOS technology. However, there are few reviews that systematically summarized the physical vapor deposition of 2D bismuth, which brings us to the purpose of this review. Several typical physical vapor deposition methods to prepare 2D bismuth will be discussed in detail below.

Under ultra-high vacuum (~10^–8 Pa) and small mismatch, atoms arrange and form high-quality single crystal films with controllable thickness, the surface being very smooth with roughness lower than 1 nm, even picometer level^{[17, 53-56]}. This enables the advantage of realizing ultra-thin bismuthene with great crystallinity and high conductivity for the MBE method.

The interaction between the substrate and 2D bismuth has a crucial effect on film growth, and substrate-dependent electronic properties and orientations on various substrates are extensively studied, e.g. SiC^[17], Au^[56], graphene^{[47, 48]}, Si^[19], Bi₂Te₃^[57], and highly oriented pyrolytic graphite (HOPG)^[58]. A scanning tunneling microscopy (STM) overview demonstrated the flat honeycomb monolayer bismuthene as a result of strain on insulate SiC (0001) structure in Figs. 3(a) and 3(b)^[17]. The substrate temperature was controlled to ~500 °C, to condense the bismuth layer on the surface. Then post-annealing at ~400 °C was performed to improve the order of the honeycomb film. Actually, the SiC substrate forms a covalent bonding with bismuthene, generating a large topological band gap of 0.8 eV at room temperature. In contrast, the work function difference in the Au (111) substrate (5.3 eV) and Bi (4.3 eV) facilitates electron transferring from bismuthene to the Au (111) substrate, and forms Van der Waals bonding in Figs. 3(c) and 3(d)^[56]. Hu et al. found that metallic Bi nanoflakes and nanorods prefer to epitaxy on the SiC region at low coverage stage as Fig. 3(e), but semiconducting Bi nanoribbons (bandgap of 0.5 eV) with several monolayers are formed on the epitaxial graphene (EG) region, indicating that the electronic states of Bi are substrate-dependent^[47]. Fig. 3(f) shows the atomic structure of Bi (110) nanoribbon on the EG layer. Similarly, an asymmetric narrow gap (~0.4 eV) is observed in Bi nanoribbons on epitaxial graphene due to quantum and size effects and the easy electronic transfer from the substrate to Bi^[48]. It is worth mentioning that (110) orientation is thermodynamically favored for epitaxial bismuthene at the initial growth stage with a thickness of several monolayers, while (111) orientation is preferred for thicker film because of the higher cohesive energy relative to (110)^[59-61]. However, (111)-orientated films can also be directly prepared by altering the substrate and temperature^{[57, 58]}.

Furthermore, Meyer et al. reported a sudden and massive generation of misfit dislocations at a critical thickness of 4 nm in Bi (111)/Si (001) for strain relaxation in Fig. 3(g)^[55]. This finding may also have implications on other epitaxial hetero-films to explore the strain state. Walker et al. transferred epitaxial 2D bismuth for the first time, making it realizable to measure the characteristics of bismuth films on transparent, flexible and insulating substrates^[53].

Nevertheless, the MBE method has a high requirement on facility with high cost and a long experimentation period, preventing it from large-scale 2D bismuth preparation.

The PLD method uses a high energy laser to vaporize and dissociate the material under 10^–5–10^–6 Pa. High ion energy promotes the adatom mobility^[62], and results in smoother films than other low energy methods (evaporation or sputtering). Actually, the surface roughness of PLD 2D bismuth is several nanometers or lower than 1 nm^[62-64], paving a way for fabricating smooth surface, high electrical conductivity and good crystallinity 2D bismuth.

An unusual epitaxial growth occuring on Bi/Si (100) at the beginning of the PLD process was first reported in 1999^[63]. Strong texture morphology in Fig. 4(a) and more rapid growth on Si (100) than on glass are observed by SEM and in-situ electrical resistance measurements respectively. When the thickness is less than 25 nm, the 2D bismuth deposited on the glass substrate is amorphous, while the one deposited on Si (100) has perfect crystallinity. Interestingly, in Fig. 4(b), (110) orientation is dominated below 22 nm, but transforms to (111) orientation in thicker films on Si (100). This thickness-dependent preferred orientation transforming behavior is similar to MBE prepared films as mentioned in 3.1 but with a larger critical thickness and has rarely been reported. These are explained to be epitaxial growth on Si (100) at the beginning growth regardless of misfit of 17%, which suggests that PLD could lead to something interesting that is worth studying. Note that the deposition rate is lower than 0.1 Å/pulse (or 0.5 Å/s), which may be responsible for the unusual epitaxy as the author suggested.

Substrate temperature and laser energy seriously affect the roughness, grain size and orientation of PLD 2D bismuth^{[25, 63, 64]}. As the substrate temperature decreases, both the grain size and the film roughness are largely decreased. Hence, highly (111)-preferred orientation bismuth films are obtained at an optimized temperature of −30 °C^[64].

Moreover, the low ionic energy (~110 eV) deposited samples have a pure (111) orientation in Fig. 4(c), while the high ionic energy (~270 eV) presented a (110) preferential orientation without substrate heating as Rodil reported^[62]. It is claimed that low surface energy (111) planes are preferred when limiting the energy supplied to the growth process unless enough energy is supplied by substrate heating or high energy bombardment. Another study reported centimeter scale (110) and (111) oriented 2D bismuth on SiO₂/Si and Al₂O₃ substrates as shown in Figs. 4(d) and 4(e) at room temperature and 100 °C respectively. The deposition process is at a rate of ~0.2 Å/pulse^[25], while Bi ion energy remaining unknown. It is obvious that the ion energy and substrate temperature of PLD have a great influence on the growth of 2D bismuth.

Compared with MBE, PLD has advantages of high deposition rates, low temperature, low cost and unlimited targets, whereas it has not demonstrated few-layer or even bilayer bismuthene yet.

In addition to a high energy pulse laser, the electron beam can also be focused onto the surface of source material in an e-beam evaporation method. Only a small part of the source material is heated by the precisely positioned electron beam, which can minimize the evaporation of crucible materials or other possible contaminations. Precisely controlled temperature and rate allow a convenient control on thickness and properties of grown 2D bismuth.

Jankowski et al. reported the controllable growth of 4–20 nm 2D bismuth films with a deposition rate of ~0.02 Å/s on α-Al₂O₃(0001) insulating substrate by electron beam evaporation^[65]. At temperature as low as 40 K, atomic diffusion is limited by kinetics, leading to high nucleation density of Bi island, then pseudo-cubic (110)-oriented Bi films are formed, which are stable up to 400 K. After annealing around 400 K, a competition of orientation growth between Bi (110) and Bi (111) is observed in Fig. 5(a). Above 450 K, ultra-smooth Bi (111) films are obtained directly with a lattice mismatch of 4.6% on α-Al₂O₃(0001).

For room temperature-deposited bismuth, Bi (110) domains are grown within the first 4 nm, followed by Bi (111) domains starting around 6 nm in Fig. 5(b)^[65]. This is somewhat similar to the epitaxial 2D bismuth in which (110) is preferred at initial stage, and transform to (111) with additional thickness^[59]. What they have in common is the low deposition rate which may be the precondition of commensurate growth of 2D bismuth in e-beam evaporation.

In contrast, Rodil et al. deposited ~180 nm Bi (111) thin films on not deliberately heated glass substrates at a high rate of ~18 Å/s by e-beam evaporation, different from low rate growth of Jankowski, and (110) orientation shows up as the thickness increases as illustrated in Fig. 5(c)^[62]. It is emphasized that during the evaporation process, a low energetic deposition growth condition, atoms are usually arriving at the substrate with energy between 0.1 and 5 eV, thus less energetic Bi atoms form (111)-oriented films at low thickness, whereas the structure changed to a more random orientation as the thickness increased because the naturally increasing substrate temperature enhanced the atomic energy as indicated in Section 3.2. The bismuth thin films deposited on Si (100) by e-beam evaporation at room temperature are also consistent with this view^[66].

Generally, e-beam evaporation provides a cost-effective scalable method to prepare 2D bismuth for device applications. It needs further improvements on surface roughness and precise control for few-layer bismuthene.

In view of the low melting point of bismuth (271 °C), thermal evaporation is suitable for deposition of 2D bismuth^{[24, 62, 67-69]}. In addition to the conventional thermal evaporation under vacuum conditions of ~10^–4 Pa, there is also something new. Lu et al. prepared 2D bismuth with a thickness of 13 nm by heating Bi powder in a quartz tube furnace, which can be seen as the thermal evaporation method as in Figs. 6(a) and 6(b)^[24]. At heating the temperature of 510 °C and carrier gas of N₂ under ~1 Pa, 300 nm SiO₂/Si substrates were placed 15.5−17 cm away from the heating center downstream, resulting in large-area continuous (110) orientation-dominated films. This report provides a low-cost method for the preparation of high-quality 2D bismuth. Compared to normal deposition, oblique angle deposition can reduce the size of crystallites, enhance the surface roughness and yield strongly hydrophobic bismuth thin films with a larger contact angle, which are promising in micro-channels to facilitate liquid flow for cooling in high-speed electronic devices^[67]. However, undesirable large surface roughness ranging from several to tens of nanometers is a drawback for the application of CMOS which require a smooth surface.

The magnetron sputtering method can achieve a high deposition rate under base pressure of 10^–4–10^–5 Pa, which is suitable for depositing bismuth thin films with thickness of hundreds to thousands of nanometers^{[62, 70, 72-74]}. Under the base pressure of 8 × 10⁻⁵ Pa and the working pressure of 2.3 Pa, bismuth thin film was deposited onto the glass substrates by the magnetron sputtering method. The results showed that 160 °C is a main watershed for morphology and growth mechanism of the bismuth thin films, as shown in Fig. 6(c)^[70]. Large columnar grains produced by high substrate temperature lead to an increasing value of the surface roughness as shown in Fig. 6(d) (tens to hundreds of nanometers)^{[62, 72, 74]}, not as smooth as MBE, PLD, e-beam evaporation mentioned above, thus, worse crystallinity is inevitable. For film orientation, most of the magnetron sputtered and thermal evaporated results mentioned above are consistent with Rodil’s view, e.g. (111) is preferred at low temperature and (110) is preferred at elevated temperature. Similar to evaporation methods, sputtered 2D bismuth needs further optimization in terms of surface roughness for potential CMOS applications.

Flash vaporization sends fine powders bit by bit to the high temperature evaporation source where powders can realize complete evaporation in a short time. Therefore, it is easy to obtain films with the same composition as the source material.

Polycrystalline bismuth thin films with thickness of 40–160 nm was prepared via flash evaporation on preheated (453 K) glass substrates under a vacuum of 2 × 10⁻⁶ Torr for the first time^[71]. Interestingly, electrical resistivity, electron concentration and mobility oscillate with increasing thickness in Figs. 6(e) and 6(f), as a consequence of the periodic step-like shape of the electron and hole density of states. Negative Hall coefficients show n-type conduction. The low electrical resistivity (300 K) in the range of 0.36–0.46 mΩ cm and high electron mobility in the range of (2.7–5.0) × 10² cm² V^–1 s^–1 prove that high quality bismuth thin films can be prepared by flash evaporation. Recently, flash vaporization is also used in the preparation of bismuth compounds^{[75, 76]} due to its controllable stoichiometric ratio.

Regarding the orientation of physical vapor deposited films, it can be summarized as follows: 1) for epitaxial 2D bismuth, the puckered-layered (110) phase is favored due to the higher cohesive energy for small thickness films. As thickness increasing, the situation is reversed, resulting in (111) dominated (Fig. 7). The critical thickness is 4–6 layers according to Nagao et al.^{[59, 60]}. However, transition from the (110) to (111) phase can be controlled by altering the deposition temperature and substrate, making it possible to directly control the crystal orientation of 2D bismuth^{[42, 58, 77]}. 2) For other physical vapor-deposited 2D bismuth including pulsed laser deposition, e-beam evaporation, thermal evaporation and magetronsputtering, a similar (110) to (111) transition along the increasing thickness would occur when the deposition rate is low enough^{[25, 63, 65]}. Furthermore, at high deposition rate, there is a competitive growth of different orientation as Rodil reported: the low surface energy (111) plane is favored and the film tends to be randomly oriented or pure (110) oriented^{[62, 65, 66, 68, 70]}, unless enough energy is supplied either by substrate heating or high energy bombardment. Even so, there is still no definite conclusion about 2D bismuth orientation and (111) to (110) transition up to now, thus more research on the growth mechanism is needed.

Above-mentioned physical vapor deposition methods can realize large area continuous 2D bismuth (methods and corresponding parameters are summarized in Table 1.), which are also easy to control the size and thickness, thus, providing platforms to obtain 2D bismuth compatible with CMOS technology. Still, more attention needs to be focused on growth mechanisms since there is still a challenge to balance between low cost and high quality. In short, the MBE method has the greatest advantages in quality and controllability, while the high cost limits the possibility of large-scale preparation for CMOS applications at present. Magnetron sputtering and thermal evaporation are common methods with low cost systems, but for the deposition of ultra-thin 2D bismuth for nanoelectronic devices, the shortcomings in crystallinity and roughness are obvious. In comparison, e-beam evaporation and PLD are not as expensive as MBE, and the products are relatively smooth and high-quality films with controllable thickness, thus holding great potential for the realizing of large area 2D bismuth for device integration in the future.

2D bismuth has been explored in a variety of CMOS relevant applications, due to good air stability and outstanding electronic properties. Below is a survey on theoretical and experimental implementations about 2D bismuth devices including field-effect transistors, sensors, photodetectors, optical devices, thermoelectric devices, topological and spintronic applications, magnetic devices and memory devices.

A field-effect transistor is an essential building block nowadays of CMOS devices. Modern electronic chips contain billions of FETs per square millimeter^{[78, 79]}. Although there have been tons of FET work on graphene, black phosphorus, TMDs, etc., 2D bismuth FET is still in its infancy^[80]. A back-gate FET with a 10-nm-thick bismuth film channel (channel length = 30 μm, width = 500 μm) on Si wafer capped with a 300-nm-thick SiO₂ layer has been fabricated^[25]. Figs. 8(a) and 8(b) show the schematic and photograph of Bi FET with 80-nm-thick Au layer as the source/drain electrodes. The output characteristics shown in Fig. 8(c) reveals a good ohmic contact between Bi and Au. The transport curve presented in Fig. 8(d) suggests a p-type channel which is Bi thin film and the field effect mobility was obtained as 220 cm² V^–1 s^–1, which is much larger than chemical vapor deposition (CVD) MoS₂ (3.6 cm² V^–1 s^–1)^[81], PLD BP (10 cm² V^–1 s^–1)^[82] and higher than silicene (100–200 cm² V^–1 s^–1)^{[83, 84]}. Nonetheless, the on/off ratio of 2D bismuth FET is quite low (<10), which could be explained by the feature of semimetal. Thus, on the basis of high carrier mobility, 2D bismuth is expected to be used in high-frequency devices like BP, and showcase the potential in spintronic application.

Because of the large specific surface area and its conductance changing with the extent of surface adsorption, 2D materials are widely utilized in sensors. Several calculation works have demonstrated that 2D bismuth is a promising candidate to detect some molecules. Bhuvaneswari et al. reported that a bismuthene nanosheet (BiNS) was calculated to show a response to the G series which is a kind of nerve agents, revealing the application in biosensors^[85]. After the adsorption of the G series, alteration in the band gap was observed, leading to change of conductivity. Princy Maria et al. used BiNS to detect the hazardous vapor molecules benzyl chloride and chlorobenzene by the transfer of charge, band gap and absorption energies based on the density functional theory (DFT) analysis, suggesting its prospects in chemical sensor^[86]. Bismuthene nanotube was studied by Snehha et al. to work as a gas sensor material for trace amounts of NH₃, NO₂ and PH₃, using the Perdew–Burke–Ernzerhof (PBE) function^[87]. Although theoretical research has been carried out intensively, there is a lack of experimental reports on 2D bismuth based sensor. Since the excellent sensing behavior predicted by DFT and PBE analysis, 2D bismuth chemi-resistor sensors will emerge in the next few years.

Bi is favored for ultra-broadband and high-responsive photodetectors due to its metallic surface state and small bulk gap. Yao et al. used PLD-grown Bi thin film to prepare a photodetector shown in Fig. 9(a)^[88]. An obvious switching behavior (Fig. 9(b)) was observed with a rather prompt response (0.9 s) and decay time (1.9 s). A high responsibility (c.a. 250 mA/W) was extracted^[88], which is much larger than graphene and MoS₂. Nonetheless, the calculated photosensitivity is quite low (0.02 cm²/W). Fig. 9(c) reveals that the Bi-photodetector presents an ultra-broadband range response from ultraviolet (370 nm) to infrared (1550 nm)^[88]. The mechanism of the Bi photodetector was shown in Fig. 9(d). The laser leads to the generation of electron–hole pairs and then move to the electrodes under the action of an external electric field to a photocurrent. Besides, a high-performance self-powered flexible photodetector based on CVD-Bi/polyimide (see Fig. 9(e)) was fabricated by Zhou et al.^[89]. It has not only fast rise and decay time but also durability and reproducibility after bending various radii of curvature and over 100 folding cycles. The time-dependent photocurrent was presented in Fig. 9(f).

Bismuthene was proved to be applied in nonlinear optical applications. Lu et al. characterized its nonlinear optical response at the visible band by Z-scan and cross-phase modulation (XPM) methods^[90]. They utilized the XPM effect in bismuthene to achieve all optical switching in spatial domain. Figs. 10(a)–10(c) shows the formation process of 532 nm laser of all-optical switching based on XPM, the changes of switching light leading to that of rings of signal light. Fig. 10(d) is the schematic experimental setup of XPM^[90]. In addition to optical switch, Chai et al. demonstrated a dissipative soliton ytterbium-doped mode-locked fiber laser at 1 μm regime with a bismuthene saturable absorber (SA) by using evanescent field interaction for the first time^[91]. The nonlinear transmittance of the SA (Fig. 10(e)) indicates saturation intensity and modulation depth are about 13 MW/cm²and 2.2%^[91]. A robust ultrafast pulse generation in the communications-band based on few-layer bismuthene has also been reported^[94]. Nevertheless, the bismuthene above is prepared by sonochemical exfoliation, while optical properties of physical vapor deposited 2D bismuth have rarely been investigated.

As a post-transition metal with a carrier mobility of up to ~ 20 000 cm²V^–1s^–1[26], bismuth has the lowest metal thermal conductivity except mercury of 10 W m^–1 K^{–1[95, 96]}. Given the mean free path of phonons is longer than electrons, the properties of bismuth are very conducive to the thermoelectric performance that can convert heat into electricity. The efficiency of conversion is determined by the dimensionless figure of merit (zT), as , where S stands for the Seebeck coefficient, σ denotes the electrical conductivity, κ represents the thermal conductivity and T is the temperature. It is worth noting that there is a gap in the thermoelectric performance of bismuth, compared with commercial thermoelectric materials like Bi₂Te₃, due to the much smaller Seebeck coefficient^[97-100].

Fortunately, according to the quantum confinement effect of Hicks & Dresselhaus ^{[92, 101, 102]}, bismuthene with reduced thickness would overturn this situation in terms of minimizing thermal conductivity while increasing the Seebeckcoefficient and keeping high electrical conductivity. For bismuth, it is predicted to show the quantum confinement effect below ~30 nm (Fig. 10(f)). As the thickness of the film is further reduced to less than 1 nm, zT value of bismuthene can exceed to 8 (zT up to 1 is expected for commercial use). Based on the quantum confinement effect, some researchers have focused on the thermoelectric properties of bismuthene theoretically in recent years. Wu et al.^[103] used the first-principles and Boltzmann transport theory to calculate the Seebeck coefficient of the single-layer bismuthene (~ 342 μV/K). In another calculation, the Seebeck coefficient of single-layer bismuthene can reach up to ~ 320 μV/K and the zT value can be 2.4 at room temperature^[93] (Fig. 10(g)). These theoretical predictions supported the above theories and indicated that reducing the thickness of 2D bismuth to a certain extent could improve the Seebeck coefficient, making it an ideal material for applications in thermoelectric devices. But in fact, due to the potential influence of surface charge and substrate coupling effect, the actual critical thickness may deviate from the theoretical value according to quantum confinement effect.

Recently, Yang et al.^[25] used pulsed laser deposition to grow 2D bismuth with the thickness of less than 30 nm on the SiO₂ substrate, but they did not report the thermoelectric properties of the bismuth film. Xu et al.^[104] theoretically discussed the thermoelectric properties of topological insulators (TIs) materials and showed that the thermoelectric properties of TIs depend on the geometric size, which provides a new idea for improving the zT values of TIs like bismuthene. Actually, as early as the 1990s, Das et al.^[105] and Cho et al.^[106] measured the Seebeck coefficient of bismuth thin films at room temperature (~ 60 μV/K). However, experimental results of thermoelectric properties, such as zT values of 2D bismuth are still lacking. So far, no one has reported the zT value of the 2D bismuth below 30 nm.

There are still great challenges and prospects for further research on the thermoelectric properties of 2D bismuth in the future, such as avoiding the oxidation during growth, controlling the surface roughness of thin films and unifying the instrument standard for the thermoelectric measurement of 2D materials.

Koroteev et al. found strong spin-orbit splitting on Bi surfaces^[16], suggesting the spintronic applications of 2D bismuth. Bi has high magnetic susceptibility, resistivity, Hall coefficient and low heat conductivity. Thus, 2D bismuth became a candidate for the topological insulator (TI). Bi (111) is a honeycomb-like hexagonal lattice, so its surface electronic state has the energy band structure of a multi-energy valley with a vortex-like spin state^[107]. Combining first-principles calculations and STM/STS experimental studies, Lu et al. reported the nontrivial 2D TI phases in 2-monolayer (2-ML) and 4-ML Bi (110) films with large and tunable bandgaps determined by atomic buckling of Bi (110) films. Their topological property is sensitive to atomic buckling which is sensitive to charge doping and could be controlled by different substrates^[36]. Du et al. reported the determination of surface-state Landau levels (see Fig. 11(a)) in deposited Bi (111) film under different magnetic fields via STM and the identification of some specific surface spin states with a large g-factor^[107]. Their findings throw light on the applications of 2D bismuth in topological devices and spintronics.

Quantum spin Hall (QSH) material has edge conductance channels, which can prevent some types of scattering, and is a hope for a revolutionary device without loss of spin current^[17]. Reis et al. fabricated bismuthene on SiC (0001) via MBE (as shown in Fig. 3(b)) for TI investigation^[17]. Bismuthene was detected to have a conductive edge state consistent with theory. They used ARPES to test the topological boundary states, deriving a large band splitting of ~0.43 eV (Fig. 11(b)). And the STS result shown in Fig. 11(c) reveals the existence of bismuthene edges at SiC-substrate steps and a large bulk gap of ~800 meV^[17]. Besides, it is exciting that Gou et al. reported the successful growth of a bismuth homostructure^[108] consisting of monolayer bismuthene and single-layer black phosphorus-like Bi on the HOPG surface as shown in Fig. 11(d). The formation of moiré superstructures (Fig. 11(e)) improves the decoupling between layers, preserving the topological properties of monolayer bismuthene. On the other hand, it also brings the modulation of topological edge states of bismuthene (Fig. 11(f)). Their work suggests a promising method to tailor the topological states by interfacial interactions.

It is inspiring to apply 2D bismuth to spintronic and topological devices, which can reduce power consumption and increase the running speed of CMOS.

High-performance magnetic and current sensors are widely used in integrated circuits and magnetoresistance (MR) materials are promising in this field. The giant magnetoresistance (GMR) head is more sensitive than the MR head. In other words, the same magnetic field change can cause a greater resistance value change for GMR head. It can achieve a higher storage density, holding a great prospect in CMOS. The following work reveals the giant magnetoresistance property of 2D bismuth and its potential applications in resistive memories.

Yang et al. fabricated 1–20 μm thick single crystal bismuth thin film by electrodeposition and suitable annealing and observed the MR up to 250% at 300 K and 380 000% at 5 K^[109], buried for the giant magnetoresistance of 2D bismuth. Wang et al. characterized that MR of MBE prepared polycrystalline 29–193 nm thick bismuth thin film was ~25% at 230 K and ~325% at 125 K in 9 T magnetic field^[110], respectively (see Figs. 12(a) and 12(b)). It is attributed to the compensation of electron and hole when the magnetic field is applied at the electron–hole resonance. Wang et al.^[111] investigated the temperature dependence of R&MR and magnetic field dependence of MR of MBE-Bi film with different sample tilt angleθ at 300 K (see Figs. 12(c) and 12(d)). The MR reached ~320% at ~130 K, demonstrating that low-cost polycrystalline Bi film can achieve large MR.

Xu et al. reported the reversible and nonvolatile tuning of electronic transport properties of the Bi-based heterostructures^[112]. Bismuth thin film (85 nm) was grown on (111)-oriented 0.7Pb(Mg_1/3Nb_2/3)O₃–0.3PbTiO₃ (PMN–PT) single-crystal substrates by PLD. Non-volatile resistance modulation and multiple resistance state can be achieved by polarization switching, appropriate electric fields applying and magnetic field controlling.

To sum up, 2D bismuth has been studied in many aspects of the CMOS field including high mobility FET, ultra-broadband and high-responsive photodetectors, optical, thermoelectric and topological devices. But there is a lack of experimental study on sensors and spintronic devices, urgently needing further research. Besides, the giant magnetoresistance of bismuth will become a fascinating application in the near future. It can be optimistically predicted that 2D bismuth will hold a bright prospect in CMOS.

Even though 2D bismuth has many peculiar properties, there will still be limitations for an elemental 2D material to realize various applications. Many strategies can be performed to enrich the properties and applications of 2D bismuth including doping, defect, strain, alloying and so on. The alloying method can be viewed as doping new atoms into Bi crystal^{[4, 113, 114]}, such as Bi_1–xSb_x, Bi₂Te₃ and Bi₂Se₃.

Bi and Sb share the same rhombohedral structure, so they could form a continuous solid solution, namely Bi_1–xSb_x. Bulk Bi_1–xSb_x is a small gap topological insulator (bandgap < 20 meV) when 0.07 < x< 0.22 whereas a semimetal whenx< 0.07 orx> 0.22^{[115, 116]}. This composition dependent bandgap will bring a new degree of freedom to adjust properties. Actually, a band gap of 35 meV at x= 0.15 for 90 nm Bi_1–xSb_x films^[115] has been reported by Ueda et al. The conductivity of 2D Bi_1–xSb_x exceeds 10⁵ Ω^–1m^–1 which is higher than that of 2D bismuth. And the metallic surface states demonstrate the potential of Bi_1–xSb_x as a spin Hall material with high electrical conductivity and possibly large spin Hall angle for spintronic applications. Additionally, semiconducting Bi-rich 2D Bi_1–xSb_x is a promising thermoelectric material at low temperature^[116-118]. The substitution of Bi by Sb atom in β-BiSb monolayer will decrease phonon group velocities and phonon lifetimes, which fundamentally leads to the decreased lattice thermal conductivity (0.55 W m^–1K^–1) than that of β-Bismuthene monolayer (1.23 W m^–1K^–1)^{[40, 119]} and an improved zT value (0.82 and 0.77 for electrons and holes, respectively).

Bismuth telluride (Bi₂Te₃) or bismuth selenide (Bi₂Se₃) have attracted immense interest due to their exceptional thermoelectric and optoelectronic properties. Bi₂Te₃ and Bi₂Se₃ also possess the rhombohedral crystal structure with the space group (R m). Both crystals can be disassemble into quintuple building blocks, consisting of Te(Se)⁽¹⁾–Bi–Te(Se)⁽²⁾–Bi–Te(Se)⁽¹⁾. The intralayer atoms are tightly bounded by covalent bonds, but the interlayer coupling are weak due to the existence of the Van der Waals gap^[120].

In bulk form, Bi₂Te₃ and Bi₂Se₃ single crystal have more excellent thermoelectric performance compared with Bi. Here, we mainly focus on the figure of merit zT, which is a critical parameter to characterize the performance of thermoelectric materials. Researchers have founded that the zT values of bulk Bi₂Se₃and Bi₂Te₃ are 0.11^[121] and 0.26^[122], respectively, significantly higher than that of bulk bismuth at 0.059^[96]. Cheng et al. obtained the zT value of single-layer Bi through theoretical calculation. The maximum zT value of n-type single-layer Bi is 2.1, and the maximum zT value of p-type single-layer Bi is 2.4^[93]. In comparison, single-layer Bi₂Te₃ and Bi₂Se₃ have higher zT values caused by quantum confinement effect. Zahid et al.^[123] have reported the zT value of single-layer Bi₂Te₃ was 7.15 at 300 K. Guo et al. have reported the maximal zT value of single-layer Bi₂Se₃ is 19 at 900 K when the carrier concentration is fixed at 3.24 × 10¹⁸ cm⁻³. Taking the anisotropy into consideration, the highest zT = 25 was obtained along the xx direction^[124].

Furthermore, Bi₂Te₃ and Bi₂Se₃ are expected to be important photoelectric materials for high-performance terahertz to infrared applications, due to the small bandgap, thickness and size-dependent light absorption, and tunable surface bandgap characteristics^{[125, 126]}. We compared responsivity (R) and response time (τ) of photodetectors based on Bi, Bi₂Te₃ and Bi₂Se₃, respectively. The photodetector performance of Bi₂Te₃ flakes stripped by using the scotch tape method (R= 74.32 A/W, = 0.42 s and = 0.44 s)^[125] and Bi₂Se₃ flakes via van der Waals epitaxy (R= 2.74 A/W, = 0.54 s and = 0.47 s)^[126] are much better than that of PLD-grown Bi films (R= 250 × 10⁻³ A/W, = 0.9 s and = 1.9 s)^[88].

Bismuth derivatives indicate that composite element is an effective method to tune the properties for more versatile device performance.

In this review, CMOS technology compatible physical vapor deposition methods of 2D bismuth are summarized especially focusing on epitaxy, evaporation and sputtering as well as preferred orientation of vapor deposited films. This sort of method provides a platform for synthesizing 2D bismuth with more flexible and controllable capability. Benefiting from fascinating electrical and thermal properties, 2D bismuth has a great potential in next-generation nanodevices according to experimental and theoretical investigations, which can be applied to CMOS such as transistors, sensors, optical modulators, spintronic and memory devices. In addition, an alloying strategy (Bi–X, X = Sb, Te, Se) is mentioned for further enhanced performance. Emerging 2D Xene, like bismuthene in this review, has drawn tremendous attention in CMOS related nanodevices. With atomic precision in layer control and scalability, physical vapor deposition techniques have advantages in realization of large-area, uniform and high-quality 2D bismuth for future CMOS compatible technology.

Despite the rapid development and promising achievements of 2D bismuth in recent years, there are remaining challenges for experimental study and practical applications. First, a low-cost and high-quality deposition method should be testified for industrial production of wafer-scale 2D bismuth. The recognized high-quality molecular beam epitaxy method, is costly and time consuming, making it limited to laboratory research. Traditional evaporation or sputtering methods need to be improved to yield smoother films with better quality for CMOS devices. Many potential applications are still in the stage of theoretical simulations. In-depth experimental research on material properties and device integration are necessary to narrow the gap between theoretical and experimental research on 2D bismuth. Moreover, advanced characterization techniques for 2D materials should be developed to obtain more precise results about the growth mechanism, properties and device performance. Nano scale device integration techniques also need improvement to accommodate the smaller size and higher integration density in the CMOS technique. Rational design and modification approaches such as alloying, doping, external magnetic field or strain, play an important role in determining electronic, chemical and physical properties of 2D bismuth, which can further enhance the performance and give birth to novel device applications.

This work was supported by the National Natural Science Foundation of China (No. 51602051), Jiangsu Province Innovation Talent Program, Jiangsu Province Six-Category Talent Program (No. DZXX-011).